Where Does a Musical Phrase Begin—and Where Should It Be Going?

Building a Piano Piece · Season Three: From Score to Interpretation: Building a Performance of Your Own · Article Three

A teacher draws a long arc over the score and writes, “This is one phrase.” The student crescendos from the first note, reaches maximum loudness near the top of the arc, then becomes softer and slower at the end. The device can work, but it readily becomes a template unrelated to the particular work: every phrase acquires the same breath, arch and ending.

A phrase is not a fixed object waiting to be circled on the page. It is a relationship that a listener infers from motives, harmony, metre, voices, repetition, cadences and performed time. It may begin on an upbeat, cross a barline and remain unfinished after a written rest. A slur may concern articulation and be shorter than a complete phrase. Four bars do not naturally constitute one sentence in every work.

The performer must solve three questions. Where does this passage acquire its initial energy? Which events continue or redirect it? What constitutes an arrival rather than a printed division?

Look for evidence before drawing the arc

Phrase evidence can be divided into five kinds.

First is motive: where a rhythm or interval is proposed, repeated, expanded or answered. Second is harmony: how stability and instability alternate, and where bass motion and cadence direct attention. Third is metre: how upbeat, accent and connection across barlines produce gravity. Fourth is voice: whether a line continues elsewhere, or whether an inner part completes something while the surface melody rests. Fifth is form: whether recurring material now functions as opening, development, return or closure.

The evidence may conflict. A melody appears to pause while the bass has not arrived. A cadence completes a local unit while a larger sequence continues. A good performance need not erase these layers. It can let the listener hear a small breath within a longer direction.

Research on expressive performance often finds systematic relationships between musical structure and timing, dynamics or articulation. It also finds substantial variation among performers; no model uniquely derives every sounding event from the score. Studies of expressive-dynamics modelling and a framework for expressive parameters support the proposition that structure enters sound. They do not authorise a prescription that every phrase must crescendo and then diminuendo.

Draw a phrase route, not merely a shape

On a copy of the score or in a practice notebook, identify five positions:

1. Departure: which event initiates the relationship, and is there an upbeat preparation?
2. Continuation: what rhythm, sequence or voice prevents the music from stopping?
3. Turn: where does harmony, register, texture or motive change function?
4. Arrival: which event fulfils the present commitment?
5. After-effect: does the music release, connect, or immediately create a new direction?

This phrase-route map is more useful than a single high point because it allows several levels within one utterance. The highest note is not necessarily the goal, and the loudest chord is not necessarily the arrival. A dissonance may deserve audibility for its tension, while the real destination is its resolution; a registral summit may be a passing event above a lower harmonic goal.

Test direction through three separate parameters

Phrases commonly emerge through dynamics, timing and articulation together, but those parameters need not peak at the same place.

Record a dynamics-led version first. Keep the reference tempo relatively stable and create direction through voicing and keystroke energy. Next record a timing-led version with restrained dynamic change but small degrees of motion, dwelling and release. Third, make an articulation-led version: hold overall loudness and pulse relatively constant while changing overlap, release and the clarity of supporting voices.

After a delay, listen blindly and mark the departure, turn and arrival you can hear without the score. If only the performer knows that a place is meant to be the climax, the plan has not yet become sound. Conversely, if all three parameters are exaggerated at the same point, the syntax may be over-explained. The fewest sufficient cues are often clearer than a stack of simultaneous signals.

Distinguish breath, cadence and stop

A piano does not literally inhale, so “breath” is a structural metaphor. It may be realised by releasing the previous note, reducing overlap, changing pedal, redistributing weight or making a small timing adjustment. It need not involve adding an audible gap.

Phrase endings are particularly vulnerable to uniform treatment as slowdown plus diminuendo. Yet a cadence can be open, conclusive, abrupt, suspended or connective. If every comma becomes a full stop, large form is fragmented. If every connection is seamless, the listener loses hierarchy.

Try removing all habitual end-of-phrase ritardandi. Play with a maintained pulse and restore time only where harmony, form or genre supplies a reason. Then allow boundaries to be formed only through release and pedal. If the phrases remain audible, the previous large slowdowns may have been unnecessary.

This does not mean that phrase-final timing is wrong. It means that timing must do a particular job. Research into expressive performance describes regular phrase-related changes, but a statistical regularity across performances does not decide what this phrase in this interpretation requires.

The same melody can acquire a different route

Repeated material does not necessarily repeat its function. The first appearance may propose; the second, carrying memory of what came before, may answer. Identical notes placed in a different harmony, register, texture or formal location have a changed direction. Do not decide in advance that the second occurrence must be louder or softer. Compare the two route maps. What new background does the later departure possess? Is its arrival more final? Which accompanying voice has changed?

In contrapuntal Bach, the boundaries of a subject may be clarified by entries and countersubject relations, while heard direction still arises from several voices together. In melodic Schumann or Chopin, a long slur cannot replace harmonic and inner-voice analysis. No universal historical rule is needed here. Begin with a reliable text, then reason from relationships within the particular work.

A seven-day phrase experiment

On day one, mark routes without adding dynamic plans. On day two, play the melodic skeleton, bass and important inner voice separately, testing whether they imply the same arrival. On day three, make the dynamics-, timing- and articulation-led versions. On day four, select the solution with the fewest sufficient cues and restore full context. On day five, make a cold recording and ask a listener who is not following the score to mark perceived divisions and goals. On day six, compare that response with the route map. On day seven, test on another piano or in another room, asking whether the design depended excessively on pedal or one instrument’s colour.

Keep a small table with columns for evidence, route position, performance parameter, listening result and next revision. “Make it sing” then becomes more than an atmospheric instruction: it becomes an audible, revisable design.

When two phrase readings remain plausible

Melodic accent can imply one boundary while harmony and bass support a longer utterance. Do not immediately declare one reading wrong. Record a short-phrase version and a long-line version, then make a third that releases gently at the smaller boundary while carrying bass and pulse across it. If the third permits local breath and long direction to coexist, it may represent the layered structure better than either exclusive account. If the layers become cluttered, choose the relation that matters most at this point. Interpretation does not emphasise every analytical observation equally; it decides which relations the listener needs now.

A phrase lives not because it has been given a decorative arch, but because it proposes a relationship, carries hearing through change and fulfils—or deliberately postpones—a commitment at an appropriate place.

Phrasing is not the addition of expression to a group of notes. It lets a listener hear where direction comes from and why an arrival belongs here.

Primary sources and further reading

• Modelling expressive dynamics in classical performance
• A framework for timing, dynamics and articulation in expressive performance
• Personal taste and modelling of piano dynamics
• Timing, articulation, dynamics, pedal and timbral individuality in piano performance
• Listening responses to expressive and mechanical performance

Continue reading: Building a Piano Piece series page

If you would like to bring these ideas about listening, understanding, and practice to the keyboard, you might try ScoreFlow, an app I developed to make score reading and daily practice flow more naturally together.